RNASeq Analysis with DESeq2

0.1 Load Packages
0.2 bioMart
0.3 Read Counts from Salmon
0.4 DESeq Dataset
0.5 Annotating Samples
0.6 Annotating Transcripts
0.7 PCA on Expression Matirix
0.8 t-SNE on Expression Matirix
0.9 Sample Similarities
0.10 Differential Expressed Genes between Treatments
0.11 P-Value Distributions for Treatment
0.12 Enrichment Analysis for Different Treatments
0.13 P-Value Distributions for all Cell Lines
0.14 Enrichment Analysis for Different Treatments for all Cell Lines

0.1 Load Packages

library(DESeq2)
library(tximport)
library(biomaRt)
library(tidyverse)
library(ggplot2)
library(ggrepel)
library(readxl)
library(Rtsne)
library(pheatmap)
library(IHW)
library(dplyr)
library(grid)
library(gridExtra)
library(jyluMisc)
library(gtable)

0.2 bioMart

# ensembl.mart <- useEnsembl(biomart = "ensembl", dataset = "hsapiens_gene_ensembl")
# annotations <-
# getBM(
# attributes = c(
# 'hgnc_symbol',
# 'refseq_mrna',
# 'transcript_biotype'
# ),
# mart = ensembl.mart
# )
# annotations <- annotations %>% filter(hgnc_symbol != "" & refseq_mrna != "")

0.3 Read Counts from Salmon

# files <-
#   file.path("../data/alignments/",
#   list.files("../data/alignments"),
#   "quant.sf")
#   sample.ids <-
#   list.files("../data/alignments") %>% map(function(x)
#   gsub("_sequence", "", strsplit(x, split = "lane[[:digit:]]+")[[1]][2]))
#   names(files) <- sample.ids
#   tx2gene <- annotations %>% dplyr::select(refseq_mrna, hgnc_symbol)
#   txi.salmon <-
#   tximport(files,
#   type = "salmon",
#   tx2gene = tx2gene,
#   ignoreTxVersion = TRUE)

0.4 DESeq Dataset

# sample.table <-
#   data.frame(row.names = sample.ids,
#   fileName = list.files("../data/alignments/"))
# dds <- DESeqDataSetFromTximport(txi.salmon, sample.table, design = ~1)

0.5 Annotating Samples

# sample.list <-
#   read_xlsx("../data/metadata/sampleList.xlsx", sheet = 2)
#   plate <- sample.list$`Plate well number`
#   id <- sample.list$ID
#   sample.name <- sample.list$`Sample Name`
#   cell_line <-
#   sample.name %>% map(function(x)
#   strsplit(x, split = "_")[[1]][1])
#   cell_line <- unlist(cell_line)
#   treatment <-
#   sample.name %>% map(function(x)
#   strsplit(x, split = "_")[[1]][2])
#   treatment <- unlist(treatment)
#   treatment <- replace(treatment, treatment == "NTC-1-old", "NTC-1")
#   repl <-
#   sample.name %>% map(function(x)
#   strsplit(x, split = "_")[[1]][3])
#   repl <- unlist(repl)
#   sample.annotations <-
#   data.frame(
#   row.names = id,
#   cell_line = as.character(cell_line),
#   treatment = as.character(treatment),
#   replicate = repl,
#   plate = plate
#   )
#   write.csv(sample.annotations, file = "../data/metadata/sample_annotations.csv", row.names = T)
#   sample.annotations <-
#   sample.annotations[sample.annotations$cell_line != "OCI-AML2", ]
#   colData(dds) <-
#   cbind(colData(dds), sample.annotations[rownames(colData(dds)), ])
#   dds$lane <- c(rep("2", 30), rep("3", 24))

0.6 Annotating Transcripts

# gene.annotations <- annotations[match(rownames(dds), annotations$hgnc_symbol),]
# rownames(gene.annotations) <- rownames(dds)
# rowData(dds) <- gene.annotations
# dds <- dds[,dds@colData$cell_line != "OCI-AML2"]
# save(dds, file = "../data/RData/3RNAseq.RData")
sample.ids <- list.files("../data/alignments") %>% map(function(x) gsub("_sequence", "", strsplit(x, split ="lane[[:digit:]]+")[[1]][2]))
sample.annotations <- read.csv("../data/metadata/sample_annotations.csv", row.names = 1)
sample.ids <- unlist(sample.ids)
sample.ids <- sample.ids[sample.annotations[sample.ids, "cell_line"] != "OCI-AML2"]
sample.annotations <- sample.annotations[sample.annotations$cell_line != "OCI-AML2", ]
load("../data/RData/3RNAseq.RData")

0.7 PCA on Expression Matirix

sample.ids <- sample.ids[which(dds$treatment %in% c("AK2-3D", "NTC-1"))]
dds <- dds[, dds$treatment %in% c("AK2-3D", "NTC-1")]

dds$treatment <- droplevels(dds$treatment)
dds$cell_line <- droplevels(dds$cell_line)
dds <- estimateSizeFactors(dds)
dds <- dds[which(rowSums(counts(dds)) > 50), ]
# dds.vst <- varianceStabilizingTransformation(dds, blind = TRUE)
dds.vst <- dds
exprMat <- assay(dds.vst)
sds <- rowSds(exprMat)
exprMat <- exprMat[order(sds, decreasing = TRUE)[1:5000], ]
pcRes <- prcomp(t(exprMat), scale. = FALSE, center = TRUE)
eigs <- pcRes$sdev ^ 2
eigs <- eigs / sum(eigs)

plotTab <- data.frame(pcRes$x[, c(1, 2)])
plotTab$sampleID <- rownames(plotTab)
plotTab$cellLine <-
sample.annotations[plotTab$sampleID, "cell_line"]
plotTab$treatment <-
sample.annotations[plotTab$sampleID, "treatment"]

ggplot(plotTab,
aes(
x = PC1,
y = PC2,
label = sampleID,
color = cellLine,
shape = treatment
)) +
geom_point() +
geom_text_repel(aes(PC1, PC2, label = sampleID)) +
theme_bw()

0.8 t-SNE on Expression Matirix

tsneRun <- function(distMat,
                    perplexity = 10,
                    theta = 0,
                    max_iter = 5000) {
                    tsneRes <- Rtsne(
                    distMat,
                    perplexity = perplexity,
                    theta = theta,
                    max_iter = max_iter,
                    is_distance = TRUE,
                    dims = 2
                    )
                    tsneRes <- tsneRes$Y
                    rownames(tsneRes) <- labels(distMat)
                    colnames(tsneRes) <- c("x", "y")
                    tsneRes
                    }
                    
                    distViab <- dist(t(exprMat))
                    
                    plotTab <-
                    data.frame(tsneRun(distViab, perplexity = 2, max_iter = 10000))
                    
                    #plot
                    plotTab$sampleID <- rownames(plotTab)
                    plotTab$cellLine <-
                    sample.annotations[plotTab$sampleID, "cell_line"]
                    plotTab$treatment <-
                    sample.annotations[plotTab$sampleID, "treatment"]
                    
                    ggplot(plotTab,
                    aes(
                    x = x,
                    y = y,
                    label = sampleID,
                    color = cellLine,
                    shape = treatment
                    )) +
                    geom_point() +
                    geom_text_repel(aes(x, y, label = sampleID)) +
                    theme_bw()

0.9 Sample Similarities

colAnno <- data.frame(plotTab[,c("cellLine", "treatment")])
corMat <- cor(exprMat)
pheatmap(corMat, annotation_col = colAnno)

0.10 Differential Expressed Genes between Treatments

dds$id <- unlist(sample.ids)
dds$cell_line <- sample.annotations[dds$id, "cell_line"]
dds$treatment <- sample.annotations[dds$id, "treatment"]
dds$treatment <- droplevels(dds$treatment)
dds$cell_line <- droplevels(dds$cell_line)
design(dds) <- ~ cell_line

# ddsCell <- DESeq(dds)
# results.cell <- results(ddsCell, contrast = c("cell_line" , "Raji", "Namalwa"))
# results.df <- as.data.frame(results.cell)
# summary(results.df)
# res.05 <- results(ddsCell, alpha = 0.05)
# table(res.05$padj < 0.05)
# resIHW <- results(ddsCell, filterFun=ihw)
# summary(resIHW)

results.cell <- list()
dds$lane <- factor(dds$lane)
dds.cells <-
unique(dds$cell_line[dds$treatment %in% c("NTC-1", "AK2-3D")]) %>% map(function(x)
dds[, dds$cell_line == x]) %>% map(function(x)
x[apply(counts(x), 1, function(x)
all(x > 10)), ])

dl <- function(x) {
x$treatment <- droplevels(x$treatment)
x$lane <- droplevels(x$lane)
if (length(levels(x$lane)) > 1) {
design(x) <- ~ treatment + lane
} else{
design(x) <- ~ treatment
}
x
}
dds.cells <- dds.cells %>% map(dl)
results.cell <-
dds.cells %>% map(function(x)
DESeq(x, betaPrior = T)) %>% map(function(x)
results(x, contrast = c("treatment", "NTC-1", "AK2-3D")))
names(results.cell) <-
unique(dds$cell_line[dds$treatment %in% c("NTC-1", "AK2-3D")])
null <-
names(results.cell) %>% map(function(x)
write.csv(results.cell[[x]], paste0("../results/DEG_cell_line/", x, ".csv")))
top.genes <-
results.cell %>% map(function(df)
rownames(df[df$pvalue < 0.05 & abs(df$log2FoldChange) > 0.5, ]))

0.11 P-Value Distributions for Treatment

plotList <- list()
results.cell %>% map(function(x)
tibble(pval = x$pvalue)) %>% map(
function(x)
ggplot(x, aes(pval)) + geom_histogram(
bins = 20,
fill = "blue",
col = "blue",
alpha = 0.3
) +
theme_bw() + theme(plot.title = element_text(hjust = 0.5)) +
xlab("raw p values")
) -> plots

plotList <-
1:length(plots) %>% map(function(x)
plots[[x]] + ggtitle(sprintf(
"NTC-1 vs AK2-3D in %s", names(results.cell)[x]
)))
names(plotList) <- names(results.cell)
grid.arrange(grobs = plotList, ncol = 2)

0.12 Enrichment Analysis for Different Treatments

createInput <-
  function(DEres,
  pCut = 0.05,
  ifFDR = FALSE,
  rankBy = "stat") {
  if (ifFDR) {
  inputTab <-
  data.frame(DEres) %>% rownames_to_column(var = "symbol") %>% filter(padj <= pCut)
  } else {
  inputTab <-
  data.frame(DEres) %>% rownames_to_column(var = "symbol") %>% filter(pvalue <= pCut)
  }
  
  inputTab <-
  arrange(inputTab, pvalue) %>% filter(!duplicated(symbol)) %>% select_("symbol", rankBy) %>% data.frame(stringsAsFactors = FALSE)
  rownames(inputTab) <- inputTab$symbol
  inputTab$symbol <- NULL
  colnames(inputTab) <- "stat"
  return(inputTab)
  }
  
  gmts <-
  list(
  H = system.file("extdata", "h.all.v5.1.symbols.gmt", package = "BloodCancerMultiOmics2017"),
  C6 = system.file("extdata", "c6.all.v5.1.symbols.gmt", package =
  "BloodCancerMultiOmics2017"),
  KEGG = system.file("extdata", "c2.cp.kegg.v5.1.symbols.gmt", package = "BloodCancerMultiOmics2017")
  )
  
  
  results.cell %>% map(function(x)
  createInput(x, pCut = 0.1, ifFDR = FALSE)) %>% map(function(x)
  names(gmts) %>% map(
  function(y)
  runGSEA(
  inputTab = x,
  gmtFile = gmts[[y]],
  GSAmethod = "page"
  )
  )) -> p
  
  p %>% map(function(x)
  x[[1]]) -> res.Halmark
  p %>% map(function(x)
  x[[2]]) -> res.Oncogenetic
  p %>% map(function(x)
  x[[3]]) -> res.KEGG
  enBar.Halmark <-
  plotEnrichmentBar(res.Halmark,
  pCut = 0.05,
  ifFDR = FALSE,
  setName = "HALLMARK")
  plot(enBar.Halmark)

  enBar.Oncogenetic <-
  plotEnrichmentBar(res.Oncogenetic,
  pCut = 0.05,
  ifFDR = FALSE,
  setName = "Oncogenetic")
  plot(enBar.Oncogenetic)

  enBar.KEGG <-
  plotEnrichmentBar(res.KEGG,
  pCut = 0.05,
  ifFDR = FALSE,
  setName = "KEGG")
  plot(enBar.KEGG)

0.13 P-Value Distributions for all Cell Lines

dds <- dds[,dds$treatment %in% c("NTC-1", "AK2-3D")]
dds$treatment <- droplevels(dds$treatment)
dds$cell_line <- droplevels(dds$cell_line)
design(dds) <- ~ cell_line + treatment
dds <- DESeq(dds)
res.all <- results(dds, contrast = c("treatment", "NTC-1", "AK2-3D"))
plotTab <- tibble(pval = res.all$pvalue)
p <- ggplot(plotTab, aes(pval)) + geom_histogram(bins = 20, fill = "blue", col="blue", alpha=0.3) +
  theme_bw() + theme(plot.title = element_text(hjust =0.5)) +
  ggtitle("NTC-1 vs AK2-3D in all cell lines") +
  xlab("raw p values")
p

0.14 Enrichment Analysis for Different Treatments for all Cell Lines

inputTab <- createInput(res.all, pCut = .1, ifFDR = F)
enRes <- runGSEA(inputTab = inputTab, gmtFile = gmts$H, GSAmethod = "page")
enRes$Name <- gsub("HALLMARK_","", enRes$Name)
enRes <- list(enRes)
names(enRes) <- "NTC-1 vs AK2-3D"
enBar <- plotEnrichmentBar(enRes, pCut = 0.1, ifFDR = FALSE, setName = "HALLMARK")
plot(enBar)

inputTab <- createInput(res.all, pCut = .1, ifFDR = F)
enRes <- runGSEA(inputTab = inputTab, gmtFile = gmts$C6, GSAmethod = "page")
enRes$Name <- gsub("Oncogenetic_","", enRes$Name)
enRes <- list(enRes)
names(enRes) <- "NTC-1 vs AK2-3D"
enBar <- plotEnrichmentBar(enRes, pCut = 0.1, ifFDR = FALSE, setName = "Oncogenetic")
plot(enBar)

inputTab <- createInput(res.all, pCut = .1, ifFDR = F)
enRes <- runGSEA(inputTab = inputTab, gmtFile = gmts$C6, GSAmethod = "page")
enRes$Name <- gsub("KEGG_","", enRes$Name)
enRes <- list(enRes)
names(enRes) <- "NTC-1 vs AK2-3D"
enBar <- plotEnrichmentBar(enRes, pCut = .1, ifFDR = FALSE, setName = "KEGG")
plot(enBar)

hv.genes <- function(df, n = 500){
  sds <- apply(df, 1, sd)
  means <- apply(df, 1, mean)
  sds <- unlist(sds)
  means <- unlist(means)
  lm <- lm(sds~means)
  sm <- summary(lm)
  res <- residuals(lm)
  names(res) <- 1:length(res)
  down <- tail(order(res), n)
  down
}

hv.genes <- function(df, n = 500){
  sds <- apply(df, 1, sd)
  down <- head(order(sds), n)
  down
}

cell_lines <- unique(dds$cell_line)
heat.map <- function(cell_line){
  line.indices <- which(dds$cell_line == cell_line)
  colAnno <- data.frame(dds$treatment[line.indices])
  rownames(colAnno) <- colnames(dds)[which(dds$cell_line == cell_line)]
  print(colAnno)
  to.plot <- exprMat[hv.genes(exprMat),]
  to.plot <- t(scale(t(to.plot)))
  pheatmap(to.plot[,line.indices], annotation_col  = colAnno, main = cell_line, show_rownames = F)
}
cell_lines %>% map(heat.map)

##     dds.treatment.line.indices.
## K10                      AK2-3D
## K11                       NTC-1
## K12                      AK2-3D
## K7                        NTC-1
## K8                       AK2-3D
## K9                        NTC-1

##     dds.treatment.line.indices.
## K13                       NTC-1
## K14                      AK2-3D
## K15                       NTC-1
## K16                      AK2-3D
## K17                       NTC-1
## K18                      AK2-3D

##    dds.treatment.line.indices.
## K1                       NTC-1
## K2                      AK2-3D
## K3                       NTC-1
## K4                      AK2-3D
## K5                       NTC-1
## K6                      AK2-3D

##     dds.treatment.line.indices.
## K43                       NTC-1
## K44                      AK2-3D
## K45                       NTC-1
## K46                      AK2-3D
## K47                       NTC-1
## K48                      AK2-3D
## K49                       NTC-1
## K50                      AK2-3D
## K51                       NTC-1
## K52                      AK2-3D

##     dds.treatment.line.indices.
## K19                       NTC-1
## K20                      AK2-3D
## K21                       NTC-1
## K22                      AK2-3D
## K23                       NTC-1
## K24                      AK2-3D

##     dds.treatment.line.indices.
## K25                       NTC-1
## K26                      AK2-3D
## K27                       NTC-1
## K28                      AK2-3D
## K29                       NTC-1
## K30                      AK2-3D

##     dds.treatment.line.indices.
## K37                       NTC-1
## K38                      AK2-3D
## K39                       NTC-1
## K40                      AK2-3D
## K41                       NTC-1
## K42                      AK2-3D

## [[1]]
## 
## [[2]]
## 
## [[3]]
## 
## [[4]]
## 
## [[5]]
## 
## [[6]]
## 
## [[7]]

pointer <- function(cell_line) {
  line.indices <- which(dds$cell_line == cell_line)
  colAnno <- dds$treatment[line.indices]
  to.plot <- exprMat["AK2", line.indices]
  df <- data.frame(y = to.plot, x = factor(colAnno))
  print(df)
  ggplot(df , aes(x, y,
  fill = factor(colAnno),
  col = factor(colAnno)
  )) + geom_point() + labs(title = cell_line)
}
cell_lines %>% map(pointer)

##       y      x
## K10 247 AK2-3D
## K11 783  NTC-1
## K12 377 AK2-3D
## K7  362  NTC-1
## K8  303 AK2-3D
## K9  376  NTC-1
##       y      x
## K13 244  NTC-1
## K14 264 AK2-3D
## K15 266  NTC-1
## K16 239 AK2-3D
## K17 479  NTC-1
## K18 338 AK2-3D
##      y      x
## K1 198  NTC-1
## K2  88 AK2-3D
## K3 302  NTC-1
## K4  75 AK2-3D
## K5 199  NTC-1
## K6  55 AK2-3D
##       y      x
## K43 385  NTC-1
## K44 198 AK2-3D
## K45 307  NTC-1
## K46  86 AK2-3D
## K47 241  NTC-1
## K48 135 AK2-3D
## K49 246  NTC-1
## K50 251 AK2-3D
## K51 139  NTC-1
## K52 134 AK2-3D
##       y      x
## K19 502  NTC-1
## K20 222 AK2-3D
## K21 473  NTC-1
## K22 317 AK2-3D
## K23 441  NTC-1
## K24 205 AK2-3D
##       y      x
## K25 408  NTC-1
## K26 268 AK2-3D
## K27 381  NTC-1
## K28 264 AK2-3D
## K29 594  NTC-1
## K30 316 AK2-3D
##       y      x
## K37 237  NTC-1
## K38 217 AK2-3D
## K39 643  NTC-1
## K40 446 AK2-3D
## K41 597  NTC-1
## K42 400 AK2-3D

## [[1]]

## 
## [[2]]

## 
## [[3]]

## 
## [[4]]

## 
## [[5]]

## 
## [[6]]

## 
## [[7]]

library(Vennerable)
vennList<- Venn(top.genes[c("Z138", "Nomo1", "BL60", "THP1")])
gVenn <- compute.Venn(vennList, doWeights  = TRUE)
plot(vennList, type = "ellipses")

## Warning in compute.E4(V, doWeights): Cant do a weighted E4

RNASeq Analysis with DESeq2

6/9/2018

Contents

0.1 Load Packages

0.2 bioMart

0.3 Read Counts from Salmon

0.4 DESeq Dataset

0.5 Annotating Samples

0.6 Annotating Transcripts

0.7 PCA on Expression Matirix

0.8 t-SNE on Expression Matirix

0.9 Sample Similarities

0.10 Differential Expressed Genes between Treatments

0.11 P-Value Distributions for Treatment

0.12 Enrichment Analysis for Different Treatments

0.13 P-Value Distributions for all Cell Lines

0.14 Enrichment Analysis for Different Treatments for all Cell Lines